Event driven infrastructure
0 likes•1,195 views
The document discusses event-driven infrastructure and how infrastructure can react to different types of events. It describes how infrastructure as code tools like Puppet, Chef, and Ansible can be used to configure infrastructure. It also discusses how serverless architectures using AWS Lambda allow infrastructure to scale automatically in response to events with no administration. Finally, it considers how event-driven infrastructure affects operational practices for DevOps.
Event driven infrastructure
- 1. EVENT DRIVEN INFRASTRUCTURE DevOpsDays Melbourne, 2015 Shiva N (@shivamaan)
- 3. A BANANA WAREHOUSE AT THE ZOO
- 4. RUN A BANANA INVENTORY EVERY HOUR
- 5. GIVE ME 5 BANANAS
- 6. MONKEY TOOK 5 BANANAS Run a banana inventory Plot monkeys banana consumption pattern Analytics on monkey banana consumption across all zoos
- 8. Event Bus Event Producers Event Transport Event Consumers
- 9. - Real time (or near real time) - Push notifications - One-‐way ‘fire-‐and-‐forget’ - Immediate action at event consumer - Informational (“monkey took bananas”) and not commands (“perform banana inventory”) 5 CHARACTERISTICS
- 10. WHY? - Decouple responsibilities - Communicate by triggering events - Scale out components horizontally - Higher responsiveness - High availability & Resiliency
- 12. AUTOMATED CONFIGURATION MANAGEMENT - Puppet - Chef - Ansible - Salt - Etc.
- 13. INFRASTRUCTURE AS CODE CLOUDFORMATION + CUSTOM RESOURCES = NO LIMITS
- 14. STACKSTORM
- 16. AWS LAMBDA A whole new world
- 17. - A zero-‐administration compute platform - Stored procedures for the cloud - Function-‐as-‐a-‐Service (NodeJS/Java) - Event driven from other services - …or triggered externally and even chained WHAT IS LAMBDA?
- 18. WHAT EVENTS DOES YOUR INFRASTRUCTURE REACT TO?
- 19. SIMPLE EVENTS
- 20. SPIKY LOAD EVENTS - Pre warming (Time) - Auto scaling (Metrics) - Feed forward loops (Predictive)
- 21. COMPONENT FAILURE EVENTS - EC2 auto healing - SQS design pattern - Service discovery + Container recycling – Consul – Serf
- 23. SECURITY EVENTS - DDos Attacks - Instance compromise
- 24. STREAMING EVENTS Cloudtrail logs streaming to Cloudwatch Subscribe Kinesis stream to Cloudwatch logs Lambda func8on subscribed to Kinesis watching for pa:erns SNS no8fica8on for insecure pa:erns in Cloudtrail logs
- 25. COMPLEX EVENTS - Event Observer Pattern - Intelligent build pipelines
- 28. HOW DOES THIS AFFECT “THE DEVOPS”
- 29. OPERATIONAL LIFECYCLE – Immutable infrastructure – No patching or in place deployments
- 31. MONITORING AND METRICS – Centralized logging – Log Analytics – Build metrics, application metrics, customer experience metrics instead of CPU, memory and OS metrics – Event Storms
- 32. IS THIS A SCARY PATH?
- 33. THANK YOU